Hadronic Taus

Hadronically decaying taus are primarily reconstructed using the hadronic calorime- ter. Additionally, the ID provides important information on the number of charged

tracks in the tau decay. Taus decay hadronically roughly 65% of the time, with the remaining 35% being leptonic decays [1]. The majority (nearly 60% of the total) of the hadronic decays include one or three charged pions and zero to three neu- tral pions. The most common hadronic decay channels are π−π0ντ (25.5%), π−ντ

(10.8%), π−π0π0ντ (9.3%) andπ−π+π−ντ (9.0%). The charged pions leave tracks

in the ID which are important for tau identification. Reconstruction

The tau reconstruction and identification is seeded by the jets described in the previous section [49]. The parameter R is kept as 0.4 but only jets with pT > 10

GeV and|η|<2.5 are considered. Tau reconstruction also requires events to have a primary vertex with three or more associated tracks.

Each jet is split into two regions when considered as a tau candidate. The first is known as the core region and is defined as the volume where R <0.2. The second is the isolation region and covers the range 0.2< R <0.4.

To be considered a tau candidate track, a track must • be in the core region

• havepT >1 GeV

• have at least two pixel hits

• have five or more additional silicon hits • have|d0|<1.0 mm

• have|z0sinθ|<1.5 mm

Tau candidates require exactly one or three such tracks to be considered. Tracks which satisfy the above conditions but are located in the isolation region are also kept for use in the tau identification.

Before running the tau identification algorithm, effort is made to reconstruct neutral pions, π0, in the core region. Each core region may contain between zero and two π0. Neutral pions predominantly decay to two photons, meaning each π0 candidate is observed as two energy clusters. Energy from pile-up, the underlying event and calorimetric noise is subtracted. The properties of the energy clusters are used to calculate how likely theπ0 candidate is to be a true π0.

The ID tracks passing the requirements listed above and the highest scoring π0 candidates are passed to the tau identification software.

Identification

The primary function of the tau identification machinery is to discriminate between hadronic taus and jets. As taus are seeded by jets, the identification step is im- portant to ensure taus and jets are not misidentified as each other. As well as the jet discriminant, there is also an electron veto to identify and discard electrons misidentified as taus.

The jet discriminant software uses a boosted decision tree (BDT) to clas- sify the tau candidates. Separate BDTs are trained for one- and three-track tau candidates. The BDTs use a number of variables which will be listed here. A full explanation of the variables can be found elsewhere [49].

The discriminant variables relate to the • energy distributions in the calorimeters • proportion of the energy in the lead track • track directions

• number of tracks in the isolation region • secondary vertex properties

• invariant mass of the tracks, neutral pions and the ratio between them • number ofπ0.

Different lists of the variables are used by the one- and three-track BDTs.

The BDT has three working points corresponding to different signal efficien- cies. This analysis uses the mediumworking point, corresponding to a 60% (40%) efficiency for one (three) track taus. The main sources of inefficiency are due to track reconstruction inefficiencies and additional tracks being included from the un- derlying event.

The electron veto is also BDT-based and is designed to identify and discard electrons faking taus. Variables used by the electron veto include the transition radiation properties, angle of the track, deposits in the electromagnetic calorimeter and the calorimeter shower shape. This analysis uses thelooseelectron veto which has a efficiency for true taus of 95%. No dedicated muon veto is used in the analysis due to their negligible expected contribution. This is because muons are minimum ionising particles and therefore deposit very little energy in the calorimeters, greatly lowering the chance of misidentification.

Performance

The signal efficiency and background rejection of the tau identification machinery are important measures of performance. The cut values on the BDT jet discriminant vary as a function of tau candidatepT, chosen to maintain a constant signal efficiency

across the differentpT bins. For one track taus, the signal efficiency of the medium

working point is set to 60% [49]. This corresponds to a 95% reduction in the number of jets faking taus. The chosen signal efficiency for three track taus is 40%. This gives a considerably better background rejection with only 0.2–1% of jets passing the BDT cut, varying with thepT of the tau candidate.

The second important performance consideration is the effectiveness of the electron veto. The electron veto is only available for one track taus as these are much more likely to be faked by electrons. Thelooseworking point is used, giving a 95% signal efficiency. The number of electrons faking taus is substantially reduced by the application of the veto. Varying as a function ofη, 95–97.5% of electrons are rejected.

The tau energy resolution of the ATLAS detector has also been measured [49]. For taus with ET = 20 GeV, the resolution is typically around 20%. There

is, however, some variation depending on the tau candidate’s number of tracks and η. The resolution decreases as energy increases, to of the order 10% for taus with ET = 100 GeV.

Selection

The primary kinematic tau selections used in the analysis are pT > 20 GeV,

|ηlead track| < 2.47, a charge of ±1 and one or three tracks in the core region

(R < 0.2). In addition, taus must pass the medium selection of the tau-jet dis- criminant and thelooseelectron veto.